Knowledge Graphs
Vector search finds similar text. Knowledge graphs follow connections. When your questions require multi-hop reasoning — “who worked on projects that used the same library as the team that built the authentication service?” — a knowledge graph doesn't just outperform vector RAG, it's often the only approach that works at all. This section covers what knowledge graphs are, how to build them with modern LLM tooling, and how to use them in AI pipelines.
In This Section
What Are Knowledge Graphs
Nodes, edges, properties, ontologies — and why knowledge graphs model information differently from relational databases. Includes key graph databases and storage options.
Building Knowledge Graphs with AI
The 6-phase LLM extraction pipeline: data ingestion, entity extraction, relationship extraction, entity resolution, graph storage, and query integration. Tools: Neo4j, LangChain, LlamaIndex.
Graphify — Codebase Knowledge Graphs
Open-source tool that maps your entire codebase into a queryable knowledge graph without sending source code to any external server. 71.5× token reduction on real codebases.
Graph RAG vs Vector RAG
When to use each approach, Microsoft GraphRAG's hierarchical community detection, LazyGraphRAG cost improvements, and the hybrid pattern that combines both.